by Dávid Ujhelyi
In December 2024, the Spanish Ministry of Culture unveiled a groundbreaking legislative proposal aimed at introducing extended collective licensing (ECL) for the development of general-purpose artificial intelligence (AI) models. The “Draft Royal Decree to regulate the granting of extended collective licenses for the massive exploitation of works and other subject matter protected by intellectual property rights for the development of general-purpose artificial intelligence models”[1] represents a significant move to harmonize AI development with intellectual property (IP) rights. This initiative signals Spain’s ambition to lead in the field of AI innovation while safeguarding the interests of rightholders.[2]
Spain’s rationale for implementing an ECL framework in AI development lies in its recognition of AI as a strategic national priority.[3] AI systems, particularly those relying on large-scale machine learning models, require vast amounts of data for training. Often, this data is sourced from copyrighted materials, raising significant legal and ethical questions about the rights of content creators. Legal challenges surrounding AI and copyright, along with potential solutions and different legal approaches, have been discussed in one of our previous analyses.[4]
The introduction of ECL for AI training aligns closely with Spain’s National AI Strategy, which emphasizes fostering a robust and ethical AI ecosystem. By addressing the complex issue of licensing data for AI training, the ECL proposal supports these objectives. It enables researchers and companies to access the resources needed to develop innovative AI technologies while ensuring that rightholders receive fair compensation. This balance is critical to fostering trust and collaboration between the creative and tech sectors.
Understanding Extended Collective Licensing (ECL)
Extended Collective Licensing is a system where collective management organizations (CMOs) are authorized to grant licenses on behalf of both their members and non-members. This mechanism is particularly advantageous in situations where obtaining individual permissions from numerous rightholders is impractical or impossible. By providing a blanket license for specific uses of copyrighted works, ECL simplifies the legal framework, making it easier for users to access and utilize protected materials on a large scale.
While the concept of ECL is not new – it has been successfully implemented in various contexts throughout the world – Spain’s adoption of ECL for AI development would mean a new, previously unseen application of ECL, reflecting the need of effective licensing and the unique demands of training modern AI models.
The Spanish Proposal and its Legal Overlaps
The Spanish Ministry of Culture’s proposal specifically targets the development of general-purpose AI models, including large-scale generative AI systems. Under this framework, certified CMOs would be empowered to issue non-exclusive licenses for the reproduction of copyrighted works needed for AI training. These licenses would apply even to works whose rightholders are not members of the CMO, provided they do not explicitly opt-out.
The draft decree also introduces safeguards for rightholders. Opting out of the ECL regime ensures that their works are excluded from any licensing agreements, preserving their control over how their intellectual property is used.
The European Union’s Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC (CDSM Directive)[5] provides for text-and-data mining (TDM) exceptions. However, the general TDM exception allow for rightholders to opt-out, creating challenges for researchers and developers who may face a patchwork of permissions and exceptions. Spain’s proposed ECL system aims to address these challenges by providing a streamlined licensing framework that reduces administrative burdens and ensures fair compensation for rightholders.
While the proposal is ambitious, it has raised concerns about potential overlaps with existing legal frameworks: The proposed ECL framework does not explicitly limit its scope to works opted out of the TDM exception. This creates a potential for duplication or conflict between the two systems (the TDM exception and the ECL for AI Development). For example, developers might question whether they need an ECL license for works already covered under TDM exceptions. If the new ECL regime covers only uses that are opt-outed under the TDM exception, the practicality of the whole system could be questioned. Furthermore, if the ECL regime provides for the option for rightholders to opt-out under the extended licensing, this would create a multi-level authorization scheme: 1. Is the use covered by the TDM exception? (if yes, no license is needed), 2. Did the rightholder opt-out from the free use? (if yes, authorization is needed from the relevant CMO), 3. Did the rightholder opt-out from the ECL regime as well? (if yes, the rightholder can give authorization). Clear guidelines will be necessary to delineate the boundaries between these frameworks, ensuring that the ECL system complements rather than complicates existing legal provisions.
It is also worth mentioning that while ECL could be called the very basis of the modern license distribution system, it may introduce further administrative burdens on users and reaching truly fair remuneration/compensation for rightholders may very well be a considerable challenge, even for practised CMOs.
The Global Implications of Spain’s Initiative
Spain’s proposal has the potential to set a precedent for other countries (even outside Europe) grappling with the intersection of AI development and intellectual property rights. As AI technologies become increasingly integral to industries ranging from healthcare to entertainment, the demand for high-quality, relevant and authorized training data will only grow. By proactively addressing the licensing challenges associated with this demand, Spain positions itself as a leader in crafting policies that balance innovation with IP rights protection. The ECL framework also reflects broader European trends toward regulating AI and ensuring its ethical development.
It is also worth mentioning, that under the role of the European Council’s Presidency, Hungary posed a questionnaire to Members States, inquiring their state of play regarding AI. The summary of the Member States’ answers is publicly available.[6] While Spain’s new legislative proposal in unquestionable interesting, most of the Member States’ are on the standpoint that while the effects of the EU’s new AI Act are not clear, there is no need for national legislation. Even if the need for legislation arises, many Member States’ think that it could only be carried out by the European Commission, on the EU level. Span’s proposal seems to contradict this view.
Regardless, as Spain moves forward with this proposal, its experience will offer valuable lessons for EU Members States and other nations seeking to navigate the challenges of AI development.
[1] Available at: https://www.cultura.gob.es/en/servicios-al-ciudadano/informacion-publica/audiencia-informacion-publica/cerrados/2024/concesion-licencias-colectivas.html.
[2] Teresa Nobre: A first look at the Spanish proposal to introduce ECL for AI training. Communia, Dec 10, 2024. Avialable at: https://communia-association.org/2024/12/10/a-first-look-at-the-spanish-proposal-to-introduce-ecl-for-ai-training/#fn1 (the same article is also available at Kluwer Copyright Blog, see: https://copyrightblog.kluweriplaw.com/2024/12/11/a-first-look-at-the-spanish-proposal-to-introduce-ecl-for-ai-training/).
[3] See: https://ai-watch.ec.europa.eu/countries/spain/spain-ai-strategy-report_en.
[4] See: https://ceuli.com/navigating-the-copyright-minefield-legal-challenges-of-ai-training-and-content-use/.
[5] Available at: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32019L0790.
[6] The Summary of HUPRES is available at: https://data.consilium.europa.eu/doc/document/ST-16710-2024-REV-1/en/pdf.
